Developing AI Tools to Strengthen Clinical Reasoning in Future Healthcare Professionals

Dr. Sullivan, associate professor in the Physician Assistant Program at Touro University Illinois and an instructional faculty member in the School of Health Sciences PA programs in New York, is exploring how AI can personalize instruction while helping students develop independent thinking and clinical judgment. His work includes custom AI tutors designed to guide students through questions and clinical cases, as well as Basecamp, a platform he cofounded to support knowledge development and clinical reasoning. His work reflects Touro's emphasis on faculty-led innovation, rigorous health sciences education, and preparing students to think critically and make independent clinical decisions.

How can artificial intelligence improve medical education and patient care?

I’ve spent much of my career thinking about how people learn and develop strategies to train clinicians to become thoughtful, independent thinkers. When generative AI came along, I was fascinated not only by what it could do for us, but also by what interacting with it might do to us. In medicine, I think AI has enormous potential to help clinicians sort through information, recognize patterns, and reduce administrative work that takes time away from patients.

However, I don’t see it replacing healthcare professionals. Ideally, AI gives us more capacity to do what humans are particularly suited for, such as exercising judgment, showing empathy, communicating, and making difficult decisions in the face of uncertainty. Of course, if AI wants to take over some of our paperwork, I don’t think many clinicians will object.

How can AI personalize learning for physician assistants and other health sciences students?

AI can personalize learning for physician assistants and other health sciences students by adapting guidance, questions, and clinical scenarios to individual students' needs. It can provide additional support to students who are struggling while challenging others to apply their knowledge to more complex cases.

What excites me most is the possibility of meeting students where they are. In a classroom of 40 students or more, everyone comes in with different strengths, weaknesses, and ways of thinking. No matter how good the professor is, it’s impossible to sit beside every student at exactly the moment they get stuck.

We must be careful though that we’re not just giving students easier access to answers. Learning happens through productive struggle: trying something, getting it wrong, figuring out why, and trying again. This is useful friction. AI should behave much like a capable professor. It should know when to help, when to challenge, and occasionally when to let the student struggle a little. Its design should make students think more, not less.

How are you using AI to improve medical education at Touro?

I received the Touro AI Innovation grant and began experimenting with different ways of using generative AI in my courses. Those initial experiments helped shape a larger interest of mine: how we can design AI around the way people learn rather than use it as just another way to deliver answers.

Since then, I’ve been exploring tools that engage students through Socratic questioning and clinical problem-solving while offering different levels of support depending on how they are doing. I also apply this approach in my classroom, using custom AI tutors that I designed to encourage critical thinking and deeper understanding. These tools function much like highly capable teaching assistants. After students watch my prerecorded lectures, I lead interactive review sessions in which they work through questions and clinical cases that challenge them to apply key concepts. Students can participate by responding in person or sending private messages through Zoom, giving them different ways to engage while allowing me to gauge comprehension in real time and adapt the lesson to the needs of the class.

One of the most important things I’ve learned is that when AI helps a student may be just as important as how it helps. Sometimes, the best thing we as teachers can do is hold back and let students wrestle with a problem before turning to AI. Ultimately, students must learn to combine AI-generated insights with clinical expertise, sound judgment, and an understanding of each patient's needs. By helping faculty identify learning gaps earlier and provide targeted support, AI also has the potential to deepen understanding, strengthen academic performance, and advance educational excellence across health sciences and medical programs.

What is Basecamp and how will you evaluate its impact on student learning?

Basecamp, the platform I cofounded, was designed to support knowledge development and clinical reasoning while making aspects of the learning process more visible to students and faculty. Across my teaching at Touro, I estimate that more than 100 students have had access to or used AI-supported learning tools, including custom tutors for Medical Physiology and capstone research, as well as other interactive educational activities.

Over time, we hope to examine patterns in conceptual understanding, application, critical thinking, and engagement, and explore how these relate to academic performance and student success. The long-term goal is to identify learning challenges earlier, provide targeted support, and continually improve teaching. In many ways, I think of this as bringing a preventive medicine approach to education: rather than waiting until students struggle academically, we want to better understand their learning needs and intervene while there's still time to make a difference.

What do Touro students say about AI tutors and critical thinking?

In spring ‘26, I surveyed 23 students who used a custom AI tutor I developed for my Medical Physiology course. The feedback was encouraging: Among the 23 students surveyed 100% of respondents said the tutor helped them apply concepts rather than simply memorize facts, and 96% said it challenged their critical thinking and made their study time more efficient.

Students also reported greater confidence before exams, stronger engagement with course material, and a better understanding of difficult concepts. Ninety-one percent said the tutor increased their confidence before exams, and the same percentage reported a better understanding of challenging material. Together, these findings suggest students perceive AI-supported learning as a valuable tool for deepening understanding, strengthening critical thinking, and building confidence.

How is AI changing the way future healthcare professionals are trained and care for patients?

I think AI is going to force us to reconsider what we mean by education and how we measure student success. If a student can get a sophisticated explanation of almost any topic in seconds, our job as educators can’t simply be delivering information anymore. We need to spend more time teaching students what to do with that information.

For example, students should learn how to question information, apply it, recognize when something doesn’t make sense, and make decisions when there isn’t a perfect answer. That means more active learning, more clinical cases, and more opportunities for faculty and students to use AI together and critically evaluate its output. Future healthcare professionals will encounter AI in clinical practice whether we embrace it in the classroom or not. We need to prepare them to use these tools effectively without becoming dependent on them.

What are the benefits and risks of using AI in medical education and clinical practice?

I’m very optimistic about what AI can do, but I also think we need to be thoughtful about what we ask for it to do for us. AI can help personalize education, make expertise more accessible, and help clinicians manage overwhelming amounts of medical information.

The danger is that we can very easily become dependent on the convenience it provides. If AI is always doing the difficult thinking for us, we risk weakening exactly the skills we need to recognize when AI is wrong. This is especially important in medicine, where clinicians remain responsible for the care of their patients, regardless of whether AI tools are involved. I think the goal should be to use AI as an amplifier of human intelligence rather than a substitute for it. If we can get that balance right, the possibilities are extraordinary.